Loading Events

« All Events

  • This event has passed.

{Seminar} @ CDS: 18th August: “Change Detection for Decisive Perspectives in Computer Vision.”

Wed, 18 Aug 21 @ 11:00 AM -- 12:00 PM

Department of Computational and Data Sciences

Department Seminar


SPEAKER     :  Dr. Santosh Kumar Vipparthi IEEE (SMIEEE)

TITLE            :  “Change Detection for Decisive Perspectives in Computer Vision”

Date & Time  :   August 18, 2021, 11:00 AM.

Venue              :   Online

 

ABSTRACT

First, I will introduce myself and my research group ‘Vision Intelligence Lab (VILab)’ at MNIT Jaipur.

I will be discussing two recent works of VILab: how “Scene Independency Matters?” for moving object detection and Challenges in detecting change in Micro Expression Recognition.

Detecting a change in the video is one of the fundamental low-level tasks in many computers vision and video processing applications. The Change Detection (CD) algorithm provides the basis for several video analytics applications such as behaviour analysis, traffic monitoring, video synopsis, action recognition, anomaly detection, object tracking, micro expression recognitions, depression, pain estimation, etc. However, dynamic & complex backgrounds, illumination variations, extreme weather, variable object densities, and object shape variations, pose significant challenges to effective MOD. These challenges make it very difficult to distinguish moving objects/regions from stationary ones accurately. Similarly, the microexpressions (ME) are subtle and involuntary in nature which briefly appears on the facial regions, it triggers when a person is hiding his actual emotions in high-stake situations.

In this talk, I will cover the key developments for analysing different modalities of CD in Moving Object Detection (MOD) and ME which are computationally efficient, robust to challenging scenarios.

I will discuss our proposed work on the significance of Scene Independency Matters with deep learning solutions to alleviate some of the existing challenges in MOD. We have analysed the issue of inconsistent and unfair evaluation practices in the existing deep learning literature and present our own proposition to address this problem. Further, I will discuss a newly introduced task of moving object recognition (MOR) where we perform simultaneous localization and classification of moving objects with deep learning frameworks. Due to the lack of available benchmark datasets for analysis, our research also resulted in several new annotated video datasets.

I will discuss the experimental analysis and results of the proposed works in MOD.

Similarly, I will discuss the importance of change detection in the field of micro-expression recognition and challenges in designing the CNN networks of ME.

Further, I will discuss the experimental analysis and results of the proposed works in ME.

Finally, I will conclude my talk with future research plans.

 

BIOGRAPHY

Santosh Kumar Vipparthi, a senior member of IEEE (SMIEEE). He established Vision Intelligence Lab at MNIT with research focused on important visual perception tasks such as moving object detection, human emotion recognition, aberrant event detection, Gesture recognition, Motion analysis, image retrieval, etc. Completed research and consultancy projects from various Govt. agencies. Total 44 articles are published in IEEE Transactions and top-rated conferences in the field of computer vision. Also guided 03 PhD, 13 Master, 16 UG projects. He is currently guiding 06 PhD. 02 MTech, and 01 UG projects.

 

Host Faculty: Prof. Anirban Chakraborty

Details

Date:
Wed, 18 Aug 21
Time:
11:00 AM -- 12:00 PM

Venue

Online